This is an application in clojure which would accept a news article as input and categorise it into any of the given categories.
If running through lein use,
lein run <filename.txt>
$ java -jar newsline-0.1.0-standalone.jar [args]
-
Words like "going" do not get recognised by wordnet at all for some reason. Need to figure a way out for this.
-
So Stemming was basically the shitiest idea in the world because half the time it doesn't work correctly.
-
It seems defining wordnet words without a pos tag actually is better because more words get recognised then.
-
Ditching the stemming idea initially(since the related synsets would automatically give us stemmed words) and now have to figure how to get more words to be recognised by wordnet.
-
Clojure shares its variables across functions. wtf?
-
Using .getOffset and .getSynsetID, I can find the value of the Synset ID for each instance of the word. Now I've realised that this value is very close for words that are very closely related.
-
I can formulate a function to calculate the difference between two synset id's, which can be used to make the value of the weight function.
-
We can also use statistical method to bolster the value of recurring words so that they are given higher priority.
- Make a list of all the synset words gathered and find the synset for each.
- Need to figure out how to add activation functions for this.
- Make a Global Key List with all words and their synsetIDs along with their activation values. All words in the article get a default activation value of 1(or some other number).
- Hide the crappy warnings from the stemmer module.
Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.